arm64 pmap_fault: a local TLB invalidation suffices when enabling write access
ClosedPublic
Actions

Authored by alc on Tue, Oct 6, 6:53 PM.

Details

Reviewers

andrew
kib
markj
manu

Commits

rG7dd9a8125f70: arm64: a local TLB invalidation suffices when enabling write access

Test Plan

Over the course of a buildworld on a 16-core Cortex-A72 machine, ~78M broadcast invalidations were replaced by local ones, and there were only 13 spurious write faults.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

alc created this revision.Tue, Oct 6, 6:53 PM

Herald added a reviewer: manu. · View Herald TranscriptTue, Oct 6, 6:53 PM

Herald added a subscriber: emaste. · View Herald Transcript

alc requested review of this revision.Tue, Oct 6, 6:53 PM

pmap_enter() should similarly be using a local invalidation when it is upgrading a R/O mapping to R/W. Comparing the amd64 and arm64 pmap_enter()s is what led me to discover all of this.

P.S. I found that Linux made similar changes last November: https://lkml.org/lkml/2025/11/4/601

markj accepted this revision.Tue, Oct 6, 8:03 PM

This revision is now accepted and ready to land.Tue, Oct 6, 8:03 PM

In D60417#1385873, @alc wrote:

pmap_enter() should similarly be using a local invalidation when it is upgrading a R/O mapping to R/W. Comparing the amd64 and arm64 pmap_enter()s is what led me to discover all of this.

amd64 will do a (targeted) remote invalidation when upgrading a R/O mapping, I believe. Do you mean that it too should only issue a local invalidation in that case?

In D60417#1385920, @markj wrote:

In D60417#1385873, @alc wrote:

pmap_enter() should similarly be using a local invalidation when it is upgrading a R/O mapping to R/W. Comparing the amd64 and arm64 pmap_enter()s is what led me to discover all of this.

amd64 will do a (targeted) remote invalidation when upgrading a R/O mapping, I believe. Do you mean that it too should only issue a local invalidation in that case?

Oh, amd64 doesn't issue an invalidation at all in that case. arm64 can't get away with that but a local invalidation there makes more sense.

kib accepted this revision.Tue, Oct 6, 8:38 PM

In D60417#1385921, @markj wrote:

In D60417#1385920, @markj wrote:

In D60417#1385873, @alc wrote:

pmap_enter() should similarly be using a local invalidation when it is upgrading a R/O mapping to R/W. Comparing the amd64 and arm64 pmap_enter()s is what led me to discover all of this.

amd64 will do a (targeted) remote invalidation when upgrading a R/O mapping, I believe. Do you mean that it too should only issue a local invalidation in that case?

Oh, amd64 doesn't issue an invalidation at all in that case. arm64 can't get away with that but a local invalidation there makes more sense.